An Improved Multi-Scale Retinex Algorithm for Vehicle Shadow Elimination Based on Variational Kimmel

The vehicle shadow´s detection and elimination work basically for extracting and tracking the vehicle characteristics, and it also plays a very important role in highway video surveillance and incident detection, affecting the post-processing of video image directly, such as vehicle tracking and speed measurement. On study of Kimmel variational and multi-scale Retinex algorithm to eliminate the vehicle shadow, a Retinex method based on anisotropy edge estimation was proposed, which took shadow edge as outliers and smooth range by range. Experiment shows that the algorithm can avoid halo effect, and shadows can be removed. It indicates this algorithm can actually be applied to the highway video surveillance and incident detection system.